Dify

Dify

By @timconsidine

The Innovation Engine for GenAI Applications

2.0.46 Updated 4 Nov 2026

CloudronVersions.json:

https://communityapps.appx.uk/cloudron-dify/CloudronVersions.json

Tags

ai assistant productivity

Description

Dify for Cloudron

An open-source LLM app development platform for building AI agents, RAG pipelines, and workflow automations.

NOTE : Dify v1.17 with Agent V2 might be a breaking change if you have previous v1.16 installation

What is Dify?

Dify is a visual platform for developing applications powered by large language models. It provides:

  • Visual Workflow Builder — Drag-and-drop interface for creating AI workflows with LLM nodes, RAG pipelines, conditionals, loops, and tool integrations
  • Agent Framework — Build autonomous agents with tool use, iterative reasoning, and human-in-the-loop support
  • RAG Engine — Built-in retrieval-augmented generation with document upload, chunking strategies, and vector search (powered by pgvector)
  • Plugin Ecosystem — Extend functionality with community plugins for tools, models, and integrations
  • Multi-Model Support — Connect to OpenAI, Anthropic, Azure, local models (Ollama/LocalAI), and 50+ providers
  • Collaboration — Team workspaces, role-based access, version control, and publish/deploy workflows as APIs

Package Features

  • Self-Hosted — Full control over data, models, and infrastructure
  • PostgreSQL + pgvector — Integrated vector database for RAG workflows
  • Redis — Caching and Celery task queues
  • Plugin Marketplace — Browse and install community plugins directly from the UI
  • Code Execution — Secure sandboxed Python/JavaScript code nodes
  • Playwright Integration — Web scraping and browser automation via bundled Playwright server

What's New in v1.19.0

  • Nested Workflow Tracing — External observability traces now include the execution steps of the Workflow Tools a workflow calls, under the invoking Tool span
  • Portable App Packages — Exported apps bundle the workflows behind their Workflow Tools, including dependencies
  • Skill Hub Automation — Manage workspace Skills through the OpenAPI and difyctl commands
  • Agent File Handling — Parallel multi-file uploads, inline PDF reading, and syntax-highlighted source previews
  • Plugin Daemon 0.6.14 — Updated plugin runtime bundled with this release

System Requirements

  • Memory: 6GB minimum (8GB+ recommended for Agent V2 workflows)
  • Storage: 10GB+ for application data and uploaded documents
  • Database: PostgreSQL with pgvector (provisioned automatically via Cloudron addon)
  • Cache: Redis (provisioned automatically via Cloudron addon)

Getting Started

  1. Install from Cloudron app store
  2. Open the app and complete the setup wizard to create your admin account
  3. Configure LLM providers in Settings → Model Providers
  4. Create your first workflow in Studio → Create from Blank
  5. Test with a simple workflow: Start → LLM → End

Documentation

Support

Screenshots

Recent Changes

* upstream Dify 1.19.0 (plugin daemon 0.6.14, sandbox 0.2.15)
* refreshed base images: cloudron/python-base:3.13-20261023 and cloudron/node-base:24-20261021
* worker queue list now mirrors upstream: adds ops_trace_v2 (new trace-delivery queue), human_input_delivery and human_input_contact_sync; ops_trace is kept so queued pre-upgrade traces still drain
* IMPORTANT - back up the database before upgrading: migration 53c10126d551 enforces one tracing configuration per app/provider and deletes superseded duplicate rows (not restored if you later downgrade)
* app exports now always use package format v2 and bundle the workflows behind Workflow Tools
* nested workflow tracing: new external traces include the execution steps inside called Workflow Tools; historical traces are not backfilled
* trigger and webhook subscriptions keep the URL stored when they were created - recreate them to pick up the current app domain
* fixed: plugin virtualenvs were deleted on every start because the check compared them against the system Python instead of their own interpreter, breaking all plugins with "ModuleNotFoundError: No module named 'dify_plugin'"
* added poppler-utils (pdftoppm/pdfinfo) for the local KnowledgeFS PDF engine